MiniMax M2.5/M2.7 — GB300 NVL72 vs MI300X
Head-to-head AI inference benchmark comparison of GB300 NVL72 (NVIDIA Blackwell) and MI300X (AMD CDNA 3) on MiniMax M2.5/M2.7. Latency, throughput, and cost across LLM workloads. Use the chart controls below to switch sequences, precisions, and metrics — same interactions as the main inference chart.
Setting 44 tok/s/user as the target on MiniMax M2.5/M2.7, GB300 NVL72 produces 6594 tok/s/chip ($0.10 per million tokens) and MI300X produces 1249 ($0.21). GB300 NVL72 is 114% cheaper per token; GB300 NVL72 delivers 428% more tok/s/chip.
At 61 tok/s/user interactivity on MiniMax M2.5/M2.7, GB300 NVL72 delivers 3816 tok/s/chip at $0.17 per million tokens; MI300X delivers 779 tok/s/chip at $0.34. GB300 NVL72 is 99% cheaper per token; GB300 NVL72 delivers 390% more tok/s/chip at this point.
GB300 NVL72 posts 1610 tok/s/chip for $0.42 per million tokens at 78 tok/s/user on MiniMax M2.5/M2.7; MI300X posts 412 tok/s/chip for $0.63. GB300 NVL72 is 52% cheaper per token; GB300 NVL72 delivers 291% more tok/s/chip. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB300 NVL72:6594.1MI300X:1248.6 | GB300 NVL72:3815.9MI300X:778.8 | GB300 NVL72:1610.0MI300X:411.9 |
| Cost ($/M tok) | GB300 NVL72:$0.099MI300X:$0.212 | GB300 NVL72:$0.171MI300X:$0.341 | GB300 NVL72:$0.417MI300X:$0.633 |
| tok/s/MW | GB300 NVL72:3110424MI300X:898287 | GB300 NVL72:1799963MI300X:560298 | GB300 NVL72:759435MI300X:296322 |
| Concurrency | GB300 NVL72:~752MI300X:~63 | GB300 NVL72:~353MI300X:~26 | GB300 NVL72:~64MI300X:~11 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.